Original papers
3D characterization of crop water use and the rooting system in field agronomic research

https://doi.org/10.1016/j.compag.2022.107409Get rights and content

Highlights

  • A new method is proposed to phenotype crop root activity in the field.

  • The proposed method captures complex GxExM interactions in field crop research.

  • The new method has potential to be applied in agronomy and crop breeding programs.

Abstract

Most field crop phenotyping research has focused on the above-ground parts of crops, ignoring a “hidden half”: the rooting system and its activity. Here we propose and test a new approach to produce 3D characterizations of crop water use and root activity in large field genotype (G) by environment (E) by management (M) experimentation, using electromagnetic induction (EMI) instrument coupled with a quasi-2D inversion algorithm, and crop canopy sensing technologies. A root activity factor (R) was calculated as a function of crop water use, soil water availability, and an indicator of crop demand. We ask i) can this approach provide accurate 3D characterizations of sorghum water use and root activity?, and (ii) does the approach capture complex GxExM dynamics?. This study was conducted based on an on-farm field experiment consisting of the factorial combination of six commercial sorghum genotypes (G), three times of sowing, two levels of irrigation (E), four plant densities (M), and three replications. Two EMI surveys ten days apart were collected using a DUALEM-21S sensor. An artificial neural network (ANN) model was developed to predict 3D soil moisture (θv) using depth-specific true soil electrical conductivity (σ, mS m−1) estimated by the inversion algorithm. Crop water use between surveys was described as the difference of θv. A multispectral index derived from satellite imagery was used as a proxy for crop demand i.e., size of the crop canopy. Principal components analysis, linear mixed models, and recursive partitioning tree techniques and crop-eco-physiological principles were used to untangle complex GxExM interactions. Results indicate that 3D crop water use could be predicted with high accuracy (LCCC = 0.81) and low prediction error (RMSE = 0.03 cm3 cm−3). The calculated water use and the value of R were significantly affected by depth, crop growth stage, irrigation treatment, plant density, and their interactions. At flowering, roots were most active at 0–1.3 m under irrigation, and deeper (0.5–1.5 m) under dryland treatment. The highest water use was for three genotypes (i.e., C, E and F) grown under irrigation and high plant densities (i.e., 9 and 12 pl m-2). The smallest water use was observed under dryland treatment, particularly for two genotypes (i.e., B and C) and high plant densities. For the crops at vegetative stages, the values of water use and R were highest in the top 0.5 m of soil depth. Larger water use was observed under dryland treatment and high plant densities, while the effects of genotypes were small (not significant). We conclude that the approach provides a rapid, accurate and cost-efficient option to phenotype crop root activity i.e., water use, in large field experimentation. We also argue that the improved understanding of the crop water use dynamics can help inform optimum combinations of genotypes and management options i.e., crop designs, across contrasting environments, and help untangle complex GxExM interactions.

Introduction

Crop phenotyping techniques are rapidly advancing. Particularly with the widespread availability of proximal sensors, the use of drones, high-resolution satellite imagery, and developments in machine learning, artificial intelligence, and data integration and analysis tools such as crop simulation models (Zhang et al., 2020, Potgieter et al., 2021). However, so far, applications have involved characterizing the above-ground parts of crops, ignoring a “hidden half”: the rooting system and its activity. Here we developed and tested a new approach to produce 3D characterizations of soil crop water use and root activity in large field genotype (G) by environment (E) by management (M) experimentation.

So far, the opaque nature of the soil has limited the options for phenotyping work on root growth and activity to the use of laborious and costly root sampling methods. Research on root systems has included growing plants in chambers (Joshi et al., 2017), pots (Richard et al., 2015), intact soil cores (McNeill and Kolesik, 2004), lysimeters (Geetika et al., 2019), and rhizotrons (Atkinson et al., 2019). An important constraint of these systems is that they introduce difficult-to-control side effects that alter the growing conditions of roots and shoots (Tracy et al., 2020). In addition, when rooting systems are sampled in the field, usually a small fraction of the whole rooting system is sampled (Ordóñez et al., 2018). A large diversity of soil water monitoring approaches exist; however most approaches sample a small volume of the soil, i.e., around probes or around access tubes of neutron moisture meter. Moreover, these approaches become impractical, laborious, and cost-prohibitive, when hundreds of pots or field plots need to be assessed multiple times over the cycle of the crop. Here we claim that in field crops research, crop phenotyping or GxExM research, there is a need for fast, accurate, and repeatable 3D approaches to quantify crop water use and root activity under field conditions.

Recent, advances in geophysical techniques e.g., electrical resistance tomography (ERT), ground penetrating radar (GPR), and electromagnetic induction (EMI), offer the potential to infer crop water use within the soil profile at spatial resolutions down to a few meters (Martínez et al., 2021). ERT requires galvanic contact between electrical probes and soils, which causes disturbance of soil surface and the need for extensive electrical cabling (Huang et al., 2016). Moreover, uncertainty remains on the performance of GPR in electrically conductive media such as on heavy-clay soil. In comparison, EMI is contactless, fast, repeatable, generates spatial-dense datasets, and can be quickly applied across large spatial areas. EMI provides measures of soil apparent electrical conductivity (ECa, mS m−1) which is influenced by soil moisture (Zare et al., 2021). ECa data is also influenced by soil texture and salinity. Interestingly, the impact of temporally dependent factors (e.g., soil moisture) can be distinguished from those of temporally independent factors (e.g., texture) with time-lapse EMI measurements (Vereecken et al., 2014), providing an understanding of changes in soil moisture over time. ECa is a depth-weighted measurement that represents soil conductivity across a soil depth range (McNeill, 1980). Inversion modelling e.g., cumulative sensitivity and Maxwell equation, can then be applied to estimate a vertical profile of the true soil electrical conductivity (σ, mS m−1), and enable the prediction of soil moisture in space and depth (Zare et al., 2020). EMI was previously used to quantify crop water use and dynamics in root phenotyping studies with wheat (Whalley et al., 2017, Blanchy et al., 2020), and chickpeas (Huang et al., 2018). Here we apply similar approaches to untangle complex interactions in root water use and root activity in large on-farm genotype GxExM studies in sorghum.

The objectives of this study were to i) answer whether EMI coupled with inversion modelling, and satellite imagery can provide an approach for 3D characterizations of sorghum water use and root activity, and (ii) research whether combining 3D soil moisture dynamics mapping and simple crop-eco-physiological principles could be used to untangle complex GxExM interactions in field research.

Section snippets

GxExM study

An on-farm trial was conducted at Nangwee, QLD Australia (27°34′4.3″ S, 151°18′38.2″ E) during the 2020–2021 Southern Hemisphere summer growing season. The climate in the region is subtropical with an average 621 mm rainfall per annum and mean annual maximum and minimum temperature of 27.0 °C and 12.0 °C, respectively (Australian Government Bureau of Meteorology, 2021). The trial covered an area of ∼3.2 ha (82 m × 384 m) of a uniform black, self-mulching cracking clay, characterized as black

Model calibration and validation

Values of R2 were determined for ANN models developed using θv, estimates of σ, and soil depth with different sets of inversion parameters (Fig. 4). The largest R2 for calibration (0.78, Fig. 4a) was obtained with FS, S2 and λ = 0.3. Thus, this combination of parameters was chosen for the inversion of the ECa and the development of final ANN model. The spatial distribution of inverted σ for the two surveys was shown in Fig. S3.

The measured values were compared to the predicted θv at all the

Discussion

A new approach, in which EMI was coupled with inversion modelling, remote sensing data, and crop eco-physiological principles, was developed to produce 3D characterizations of soil crop water use and root activity in large on-farm G×M×E experimentation. Moreover, numerical and statistical techniques (i.e., PCA, LMM and recursive partitioning trees) were used to untangle complex GxExM interactions. Results indicate that the new approach could be used to produce 3D characterizations of crop water

Conclusion

This study shows for the first time how EMI coupled with inversion techniques, an ANN model, remote sensing data, and crop eco-physiological principles, can be used to characterize sorghum water use and root activity in 3D in a large field GxExM experimentation. Validation results indicated that the approach achieved a satisfactory prediction (LCCC = 0.81; RMSE = 0.03 cm3 cm-3) on the soil water dynamics and crop water use. Root activity in depth was different between irrigated and dryland

CRediT authorship contribution statement

Dongxue Zhao: Conceptualization, Methodology, Software, Formal analysis, Visualization, Writing – original draft. Joseph X. Eyre: Resources, Data curation, Writing – review & editing. Erin Wilkus: Data curation, Resources, Writing – review & editing. Peter de Voil: Resources. Ian Broad: Resources, Data curation. Daniel Rodriguez: Supervision, Conceptualization, Project administration, Writing – review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

This work is funded by the Grains Research and Development Corporation (GRDC) project UOQ1906-010RTX. The authors acknowledge Dr. Yan Zhao and A/Prof. Andries Potgieter from UQ- QAAFI for their valuable comments on processing the satellite images. The authors also want to thank the collaborating farmers for their aid in the field experiment.

References (59)

  • R.A. Ordóñez et al.

    A solution for sampling position errors in maize and soybean root mass and length estimates

    Eur. J. Agron.

    (2018)
  • M.J. Robertson et al.

    Water extraction by grain sorghum in a sub-humid environment. II. Extraction in relation to root growth

    Field Crops Res.

    (1993)
  • M. Swoish et al.

    Comparing satellites and vegetation indices for cover crop biomass estimation

    Comput. Electron. Agric.

    (2022)
  • T.R. Tenreiro et al.

    Using NDVI for the assessment of canopy cover in agricultural crops within modelling research

    Comput. Electron. Agric.

    (2021)
  • S.R. Tracy et al.

    Crop improvement from phenotyping roots: highlights reveal expanding opportunities

    Trends Plant Sci.

    (2020)
  • H. Vereecken et al.

    On the spatio-temporal dynamics of soil moisture at the field scale

    J. Hydrol.

    (2014)
  • E. Zare et al.

    Two-dimensional time-lapse imaging of soil wetting and drying cycle using EM38 data across a flood irrigation cotton field

    Agric. Water Manag.

    (2020)
  • E. Zare et al.

    Scope to map available water content using proximal sensed electromagnetic induction and gamma-ray spectrometry data

    Agric. Water Manag.

    (2021)
  • C. Zhang et al.

    High-resolution satellite imagery applications in crop phenotyping: an overview

    Comput. Electron. Agric.

    (2020)
  • D. Zhao et al.

    Mapping cation exchange capacity using a quasi-3d joint inversion of EM38 and EM31 data

    Soil Tillage Res.

    (2020)
  • C. Bai et al.

    The relationships between seedling root screens, root growth in the field and grain yield for wheat

    Plant Soil

    (2019)
  • G. Blanchy et al.

    Accounting for heterogeneity in the θ–σ relationship: Application to wheat phenotyping using EMI

    Vadose Zone J.

    (2020)
  • E. Boenecke et al.

    Determining the within-field yield variability from seasonally changing soil conditions

    Precis. Agric.

    (2018)
  • Butler, D.G., Cullis, B.R., Gilmour, A.R., Gogel, B.G., Thompson, R., 2017. ASReml-R Reference Manual Version 4. VSN...
  • Dualem Inc. DUALEM-21S User's Manual. Dualem Inc., Milton, ON, Canada...
  • D.A. Elizondo et al.

    Neural network models for predicting flowering and physiological maturity of soybean

    Trans. ASAE

    (1994)
  • EMTOMO. 2018. EM4Soil version 3.05. EMTOMO, R. Alice Cruz 4, Odivelas, Lisboa,...
  • K.G. Falk et al.

    Computer vision and machine learning enabled soybean root phenotyping pipeline

    Plant Methods

    (2020)
  • F.C. Frischknecht

    Electromagnetic physical scale modeling

    Electromagnetic Methods in Appl. Geophys.-Theory

    (1988)
  • Cited by (8)

    View all citing articles on Scopus
    View full text